In previous work in collaboration with Dr. Minoru Ko (LG-NIA),we were able to establish a stem-cell differentiation assay based on cell morphology alone, using phase-contrast imaging without specific markers. An important validation of this approach was that we were able to establish conditions where cell lines showed consistent grouping by gene function over several independent experiments. Additionally, we were able to expand this analysis to compare the morphologies of 40 cell lines. This work has been submitted for publication. In preliminary work, we have been using this assay to characterize early events in the osteogenesis/adipogenesis pathway. The balance of osteogenic/adipogenic cells changes during aging, and this differentiation pathway may play an important role in age-related bone loss as well as accumulation of fat in bone marrow. We have continued work characterizing the molecular basis of morphological age-state transitions during the C. elegans life-span. In published work we used WND-CHARM to identify distinct morphological aging states in C. elegans. We used this technique to sort worms based on their age state during a transition period where an aging population is evenly divided between individuals in Stage I and Stage II. The worms were identical genetically, by chronological age, by growth conditions, and by visual appearance, and could only be sorted into age-states using WND-CHARM. Micro-array experiments performed on these two sub-populations revealed several hundred genes with significantly altered expression profiles. By comparing our gene lists with those from other aging studies in C. elegans, we were able to identify several gene families and functional groups that were unique to our study. A prevailing theme of the aging genes uniquely identified in this study were those involved in targeted proteolysis, which appears to be a hallmark of this first aging state transition. This study appeared in the journal AGE. Currently, the bulk of our efforts are dedicated to completing the expression profiling in all states and transitions, with the expectation that a comparative analysis of gene expression patterns across several transitions will help us refine our gene selections for subsequent validation experiments on transition timing and lifespan. Recent progress includes the development and refinement of a state classifier that can identify the aging state of individual worms directly without using thresholds and cutoffs in classifier-assigned age-scores. The classifier is now also very robust against changes in experimental and imaging conditions, which was a consistent difficulty with this project. In many biological systems characterized by transiently stable states (e.g., cell cycle, apoptosis, circadian rhythms, the lysogenic/lytic cycle in phage), the mechanism at the execution point governing the transition to a new state does not involve de novo gene expression. Instead, these immediate-early genes typically accumulate during the stable phase to a threshold level or respond to a trigger and initiate the transition independently of transcription or protein synthesis. Lastly, we have begun using RNAi to investigate the effect of altering the expression patterns of the age-state-specific genes that we've found on the timing of these age-state transitions. If similar mechanisms govern aging transitions, differential expression assays bracketing the transitions will not necessarily detect the genes acting as initiating triggers, if indeed such genes exist. However, a comparison of expression profiles within a state shortly after a transition and shortly before the subsequent transition may uncover the accumulation of these triggering genes. Our ongoing work is thus to fully characterize the expression profiles both within and between all observed aging states in C. elegans. A critical difference between our approach and other studies of gene expression across lifespan is that in our case our samples for comparison are defined by observed physiological changes rather than chronology alone. In fact, in our experiments, the samples for comparing gene expression patterns are being collected specifically on transition days, when the population is mixed and enriched for late pre-transition and early post-transition individuals. Systematic comparisons of these expression profiles is expected to provide us with a refined set of candidate genes to assay in RNAi experiments for their ability to disrupt the timing of these transitions. The accumulation of many independent training sets has allowed us to develop and refine a state classifier trained on age-states rather than on the mixed populations characteristic of many of the classes defined by chronology alone. Not surprisingly, this classifier is much more robust in assigning individuals to states, because it was not trained with mixed populations, and no longer relies on computing age scores and imposing score cutoffs to define the states. The inclusion of many independent lifespan experiments in this set has also made the classifier much more robust with regard to experimental and imaging conditions. The age-state classifier clearly shows that essentially on each day of the worm's lifespan, the population consists of mixed populations. Even on days where a particular age-state is predominant, as much as 30% of the population is in other states, making assays relying on chronological age alone difficult to interpret. Sample collection of the two states at each identified transition is nearly complete. Our strategy is to collect 500 worms in each state at each of the three transitions in multiple independent aging experiments (up to five per transition). The experiment is designed to allow comparison of gene expression patterns both between states and between early post-transition and late pre-transition within each state. We will use a bioinformatics approach similar to what we have done previously to divide the set of differentially expressed genes into functional classes and then use the multiple comparisons possible with this cohort to identify functional classes of genes that are consistently accumulated within a state and then reduced in subsequent states. Our hypothesis is that this class would contain the most likely candidates for genes that regulate the state transitions. The second class of genes we are interested in consist of the functional classes that consistently change their expression patterns across a state transition. This class represents going through an aging transition as opposed to its regulation. We expect that, for example, any transcription factors we find with these expression characteristics may also potentially affect transition timing and lifespan when assayed by RNAi. The remainder of the genes in this class should provide important clues about the consequences of the aging process, potentially identifying new biochemical and genetic pathways affected by aging. We have begun to characterize the effect of RNAi knockdown on the genes identified as differentially expressed in the first age-state transition. A group of genes producing secreted proteins with cysteine-rich regions appears to contain genes that can either advance or delay the first transition. We are also looking at a set of heat-shock genes who's expression is among the most significantly altered in the first transition.